/knowledge-extract

allowed-tools: Task, Read, Write, Bash(fd:*), Bash(rg:*), Bash(jq:*), Bash(gdate:*), Bash(eza:*), Bash(bat:*)
name: "Knowledge Extract"
description: "Extract comprehensive domain knowledge and architectural patterns using parallel analysis"
author: "wcygan"
tags: ["meta","extract"]
version: "1.0.0"
created_at: "2025-07-14T00:00:00Z"
updated_at: "2025-07-14T00:00:00Z"

Context

• Session ID: !`gdate +%s%N`
• Target project: $ARGUMENTS
• Project structure: !`fd . -t d -d 3 | head -10 || echo "No directories found"`
• Build system detection: !`fd "(deno\.json|package\.json|Cargo\.toml|go\.mod|pom\.xml|build\.gradle)" . -d 3 | head -5 || echo "No build files detected"`
• Documentation exists: !`fd "(README|ARCHITECTURE|DESIGN|DOCS)" . -t f -d 2 | head -3 || echo "No docs found"`
• Code languages: !`fd "\.(rs|go|java|ts|js|py|rb)$" . | head -5 | sed 's/.*\.//' | sort -u | tr '\n' ' ' || echo "No code files"`
• Repository size: !`eza -la . | head -3 || ls -la . | head -3`

Your Task

STEP 1: Initialize knowledge extraction session with state management

• CREATE session state file: `/tmp/knowledge-extract-state-$SESSION_ID.json`
• INITIALIZE extraction scope and project boundaries
• DETERMINE primary technology stacks and architectural complexity
• ASSESS existing documentation completeness and quality gaps

echo '{' > /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "sessionId": "'$SESSION_ID'",' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "timestamp": "'$(gdate -Iseconds 2>/dev/null || date -Iseconds)'",' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "project": "'$ARGUMENTS'",' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "phase": "initialization",' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "completed_phases": [],' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "discovered_domains": [],' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "architectural_patterns": [],' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '  "knowledge_artifacts": []' >> /tmp/knowledge-extract-state-$SESSION_ID.json
echo '}' >> /tmp/knowledge-extract-state-$SESSION_ID.json

STEP 2: Parallel domain discovery using strategic sub-agent delegation

TRY:

• LAUNCH 8 parallel sub-agents for comprehensive codebase analysis
• EACH sub-agent focuses on specific architectural aspect
• COORDINATE findings through session state management
• SYNTHESIZE results for architectural understanding

**Parallel Sub-Agent Knowledge Extraction:**

LAUNCH parallel sub-agents for simultaneous domain analysis:

• **Agent 1: Domain Model Discovery**: Analyze core business entities, data models, and domain types
• Focus: struct/class/interface definitions, enums, data transfer objects
• Extract: Business terminology, entity relationships, data constraints
• Save findings: Domain entities, business rules, validation logic

• **Agent 2: Service Architecture Analysis**: Map service layers, handlers, and business logic patterns
• Focus: Service implementations, controllers, handlers, repositories
• Extract: Service boundaries, dependency patterns, integration points
• Save findings: Service interfaces, business workflows, architectural layers

• **Agent 3: API & Interface Documentation**: Discover all external and internal APIs
• Focus: REST endpoints, RPC services, GraphQL schemas, OpenAPI specs
• Extract: API contracts, request/response patterns, authentication flows
• Save findings: Endpoint inventory, API documentation, integration guides

• **Agent 4: Data Architecture Mapping**: Analyze database schemas, migrations, and data flow
• Focus: Migration files, ORM models, SQL queries, data transformations
• Extract: Schema evolution, data relationships, query patterns
• Save findings: Database design, data flow diagrams, migration strategies

• **Agent 5: Business Logic & Workflow Analysis**: Identify state machines, business rules, and processes
• Focus: State management, workflow engines, business calculations, validation rules
• Extract: Business processes, state transitions, rule engines
• Save findings: Workflow documentation, business rule catalog, process maps

• **Agent 6: Configuration & Infrastructure Discovery**: Map deployment, configuration, and operational patterns
• Focus: Config files, environment variables, deployment manifests, infrastructure as code
• Extract: Deployment patterns, configuration management, operational procedures
• Save findings: Deployment guides, configuration documentation, operational runbooks

• **Agent 7: Error Handling & Monitoring Analysis**: Document error patterns, logging, and observability
• Focus: Error definitions, logging patterns, metrics, monitoring, alerting
• Extract: Error handling strategies, observability patterns, debugging guides
• Save findings: Error catalogs, monitoring documentation, troubleshooting guides

• **Agent 8: Testing & Quality Patterns**: Analyze testing strategies, coverage, and quality patterns
• Focus: Test structure, mock patterns, integration tests, quality gates
• Extract: Testing methodologies, quality standards, coverage patterns
• Save findings: Testing guides, quality documentation, best practices

**Sub-Agent Coordination Protocol:**

• Each sub-agent executes independently using Task tool
• Results saved to session state under respective domains
• Parallel execution provides 6-8x performance improvement
• Failed agents report gracefully without blocking others
• Main agent synthesizes findings across all domains

CATCH (analysis_failed):

• LOG error details to session state
• CONTINUE with available analysis results
• DOCUMENT gaps and limitations in final report

STEP 3: Architectural synthesis and pattern identification

TRY:

• AGGREGATE findings from all sub-agents
• IDENTIFY cross-cutting architectural patterns
• MAP domain relationships and dependencies
• EXTRACT recurring design patterns and conventions